Impermanent loss in concentrated liquidity: formula and examples

Impermanent loss is the gap between what a liquidity position is worth and what the same starting tokens would be worth if you had just held them. This guide gives the formula, a table you can check, the concentrated-liquidity version with worked numbers for narrow and wide ranges, and how much in fees a position needs to break even.

lp.xyz team8 min read
RiskConcentrated liquidity

In short

Impermanent loss (IL), also called divergence loss, is the value of your LP position minus the value of simply holding the tokens you deposited, measured at today's price.

For a full-range (constant-product) position, IL = 2√r / (1 + r) − 1, where r is the new price divided by the entry price. A 2x move in either direction costs 5.72% versus holding.

Concentrated liquidity magnifies it. For the same 5% move, a range of plus or minus 50% loses about 4 times as much as full range, and a range of plus or minus 10% loses about 20 times as much.

Fees are what pay for it: a position beats holding only when the fees it earned exceed the IL, in dollars, at the moment you withdraw.

What is impermanent loss?

Impermanent loss is the difference between the value of a liquidity position and the value of holding the same starting tokens outside the pool. It happens because an AMM rebalances your deposit as price moves: when one token rises, arbitrage traders buy it from the pool, so the pool (and your share of it) ends up holding less of the winner and more of the loser.

IL is a comparison, not a balance. Your position can be up in dollars and still have IL, because holding would have been up more. Orca's impermanent loss guide makes the same point: it is a relative comparison against holding the deposited tokens, not always an absolute loss against your deposit.

What is the impermanent loss formula?

The impermanent loss formula for a full-range, 50/50 constant-product position is IL = 2√r / (1 + r) − 1, where r = new price / entry price. It comes from the x × y = k invariant: Raydium's impermanent loss page derives the ratio of LP value to hold value, 2√r / (1 + r), and subtracting 1 writes it as a negative number. The table below is computed from it.

To run your own numbers, full range or with a price range, use the impermanent loss calculator, which uses the formulas on this page.

r = 1.25 (price +25%)
IL = -0.62%
r = 1.5 (price +50%)
IL = -2.02%
r = 2 (price +100%)
IL = -5.72%
r = 3 (price +200%)
IL = -13.40%
r = 5 (price +400%)
IL = -25.46%
r = 0.5 (price -50%)
IL = -5.72%
r = 0.25 (price -75%)
IL = -20.00%

Two things stand out. First, IL depends only on the ratio, so doubling and halving cost the same (5.72%). Second, it grows slowly at first and fast later: a 25% move costs well under 1%, a 5x move costs a quarter. Only the start and end price matter, not the path, which is why a quick round trip leaves no IL.

A quick example

Say you deposit $5,000 of a token at $100 (50 tokens) and $5,000 USDC into a full-range pool, $10,000 in total. The token doubles to $200. Holding is now worth 50 × $200 + $5,000 = $15,000. The pool position is worth $10,000 × √2, about $14,142. The gap is $858, or 5.72% of the hold value. You made money, just less than holding.

Why is it called impermanent?

Impermanent loss is called impermanent because it disappears if price returns to exactly where you entered. That name is misleading in practice for three reasons.

  • It becomes permanent the moment you withdraw at a different price than you entered.
  • Prices, especially for newer tokens, often do not come back.
  • In concentrated liquidity, once price leaves your range your position stops trading and holds one token, so waiting for a return means earning no fees while you wait.

Divergence loss is the more accurate name: the loss grows with how far the two prices diverge from your entry.

How does impermanent loss work in concentrated liquidity?

Impermanent loss in concentrated liquidity is the same comparison (position versus holding the starting tokens), but a narrow range concentrates the rebalancing into a smaller price band, so the same price move costs more. Concentrated liquidity is what Uniswap v3 and v4, Orca Whirlpools, Raydium CLMM, Meteora DLMM and PancakeSwap v3 use: you choose a price range from pa to pb and your liquidity only trades while the price is inside it.

The formula for a range

The Uniswap v3 whitepaper (§6.4.1, eqs. 6.29 and 6.30) gives the token amounts x and y for a position with liquidity L at price P inside the range; the position value V(P) follows from them:

  • Base token x = L × (1/√P − 1/√pb)
  • Quote token y = L × (√P − √pa)
  • Position value V(P) = x × P + y = L × (2√P − √pa − P/√pb)

Outside the range, clamp P to the nearest edge when computing x and y: below pa the position is all base token, above pb it is all quote token. The hold value is the starting amounts priced at today's price, H(P) = x₀ × P + y₀. IL is V(P) / H(P) − 1. L cancels, so the result only depends on the range and the prices.

Worked comparison: full range vs plus or minus 50% vs plus or minus 10%

Take a token at $100 and three positions opened at that price: full range, a $50 to $150 range (plus or minus 50%), and a $90 to $110 range (plus or minus 10%). Each column is IL versus holding the same starting tokens, computed with the formulas above.

Price -5% ($95)
Full range -0.03%. $50 to $150: -0.14%. $90 to $110: -0.67%.
Price +5% ($105)
Full range -0.03%. $50 to $150: -0.13%. $90 to $110: -0.61%.
Price -10% ($90)
Full range -0.14%. $50 to $150: -0.57%. $90 to $110: -2.83% (now at the edge, all token).
Price +10% ($110)
Full range -0.11%. $50 to $150: -0.48%. $90 to $110: -2.32% (now at the edge, all USDC).
Price +25% ($125)
Full range -0.62%. $50 to $150: -2.67%. $90 to $110: -8.55%.
Price +50% ($150)
Full range -2.02%. $50 to $150: -8.89%. $90 to $110: -17.34%.
Price -50% ($50)
Full range -5.72%. $50 to $150: -22.30%. $90 to $110: -32.54%.

For small moves inside the range, the plus or minus 50% range loses about 4.2 times as much as full range and the plus or minus 10% range about 20.4 times as much. Those multiples line up with each range's capital-efficiency factor, roughly 1 / (1 − (pa/pb)^(1/4)) (exact for a range geometrically centred on the price), which is about 4.16 for $50 to $150 and 20.44 for $90 to $110. The same concentration that multiplies IL also multiplies your share of fees while price stays inside, which is the whole trade.

What happens to impermanent loss when price leaves your range?

When price leaves a concentrated range, the position stops trading and holds 100% of one token, so its token amounts are frozen while the hold portfolio keeps moving. Above the range you hold only the quote token; below it you hold only the base token. The position earns no fees until price comes back.

Using the $90 to $110 position from above with $10,000 deposited at $100 (about 47.56 tokens and $5,244.20 USDC):

  • Price exits above. At $110 the position is all USDC, worth $10,232.12, and it stays exactly that in dollars. Holding is worth $10,475.58 at $110, $10,951.16 at $120 and $11,426.74 at $130. IL widens from 2.32% to 6.57% to 10.45% as price keeps rising, because you already sold everything.
  • Price exits below. At $90 the position is all token, worth $9,255.31 versus $9,524.42 for holding (IL 2.83%). At $80 it is 9.08% and at $70 16.03%, because you are now fully exposed to the falling token while the hold portfolio still had half in USDC.

In other words, the IL is "fixed" only in token terms. The dollar gap to holding keeps changing with price, and nothing you earn offsets it until you either reposition or price returns. lp.xyz's risks and safety page calls an out-of-range position earning nothing the worst of both.

How much in fees do you need to break even on impermanent loss?

To break even, a position's fees must at least equal its IL in dollars: fees needed = hold value − position value at the moment you withdraw. Put another way, the fee income as a share of position value has to match |IL|.

Worked example, again $10,000 deposited at $100, price moves to $105:

Full range (50/50)
Position $10,246.95 vs hold $10,250.00. Fees needed: about $3.05.
$50 to $150
Fees needed: about $12.80.
$90 to $110
Position $10,175.47 vs hold $10,237.79. Fees needed: about $62.32 (0.61% of the hold value).

The narrow range needs about 20 times the fees for the same move, and while price stays inside it, it also earns roughly that much more per dollar (if other LPs are not crowding the same ticks). Whether it comes out ahead depends on how much volume trades inside your range before price leaves it. Count gas and platform fees on the cost side too: every claim and reposition is a transaction.

On lp.xyz the activity and PnL view shows a position's value and its fees separately, rather than one combined number, because fees earned are not profit until they exceed the divergence.

How can you reduce impermanent loss?

You cannot remove impermanent loss from a two-token position, but you can choose how much of it you take on. None of these guarantee a profit.

  • Use wider ranges when you will not manage the position often. Less amplification, fewer fees per dollar, and price leaves the range less often.
  • Pick correlated pairs. Stablecoin pairs and pairs like SOL and a SOL liquid-staking token move together, so r stays near 1 most of the time. They can still depeg.
  • Use single-sided walls when you have a direction. A bid or ask wall is IL by design: it converts one token into the other across a range, like a limit order. If that conversion is what you wanted, the "loss" is just the trade you placed.
  • Hedge the volatile side with a short or an options position elsewhere. That adds its own costs, funding and liquidation risk.
  • Avoid tokens you would not hold. The bigger loss for most LPs is the token itself going to zero, which no range fixes.

Related

This guide is educational, not financial advice: providing liquidity can lose money, and the numbers above are hypothetical examples.

More guides

Start provisioning.

Connect a wallet and LP across 6 AMMs on Solana and EVM from one place.